hi derek, i would recommend at this point to use the HCP pipelines. however, if you are comfortable hacking the code, the bedpostx step itself can be replaced by a corresponding call to a nipype workflow, or better still if you have a GPU to use the nipype bedpostx GPU interface.
for standard bedpostx nipype simply parallelizes the operation across slices, but is able to do so on any cluster environment (SGE/LSF/PBS/SLURM). thus, you simply reduce the time by assigning a slice to a core. how much you can parallelize depends on our cluster configuration. while we haven't done this yet, the nipype bedpostx workflow could easily be made to look exactly like the command line call, to be a drop in replacement. however, if you have an Nvidia GPU (such as a tesla or equivalent), you can drastically reduce the compute time. for our standard sequences (about 70 directions, 60 slices), we can run bedpostx for a given participant on the gpu in less than 10 minutes. cheers, satra On Mon, Feb 23, 2015 at 10:02 AM, Archer,Derek B <[email protected]> wrote: > Hi Matt, > > > > Is it possible to use nipype for bedpost processing? Will this allow me > to use the grad_dev.nii.gz file? > > > > I haven’t seen any other people using bedpost via nipype within this > forum, however, it is being recommended to use by the engineer working for > the cluster. > > > > Thanks, > > Derek > > > > *From:* Glasser, Matthew [mailto:[email protected]] > *Sent:* Tuesday, February 17, 2015 12:55 PM > > *To:* Archer,Derek B; Stamatios Sotiropoulos; > [email protected] > *Subject:* Re: [HCP-Users] Processing Diffusion MRI Data > > > > Are you running the command on ${StudyFolder}/${Subject}/T1w/Diffusion ? > > > > You’ll need the latest version of FSL 5.0.8 on your cluster as well. > > > > Peace, > > > > Matt. > > > > *From: *<Archer>, Derek B <[email protected]> > *Date: *Tuesday, February 17, 2015 at 11:53 AM > *To: *Matt Glasser <[email protected]>, Stamatios Sotiropoulos < > [email protected]>, "[email protected]" < > [email protected]> > *Subject: *RE: [HCP-Users] Processing Diffusion MRI Data > > > > Hi Matt, > > > > The exact command I’m using locally is (I tried this from Stam’s input): > > > > bedpostx <subject directory> –n 3 –model 2 –g –rician > > > > Subject directory contents: > > bvals > > bvecs > > data.nii.gz > > nodif_brain_mask.nii.gz > > grad_dev.nii.gz > > > > This works. How long do you estimate this will take? Because the –g > option is not available on the cluster I am working on (not sure why this > is the case..), and I think that may be causing the issue. > > > > Thanks, > > Derek > > *From:* Glasser, Matthew [mailto:[email protected] > <[email protected]>] > *Sent:* Tuesday, February 17, 2015 12:47 PM > *To:* Archer,Derek B; Stamatios Sotiropoulos; > [email protected] > *Subject:* Re: [HCP-Users] Processing Diffusion MRI Data > > > > How about posting the exact command line you are using and a listing of > the contents of <subject_directory> > > > > Peace, > > > > Matt. > > > > *From: *<Archer>, Derek B <[email protected]> > *Date: *Tuesday, February 17, 2015 at 9:20 AM > *To: *Stamatios Sotiropoulos <[email protected]>, " > [email protected]" <[email protected]> > *Subject: *Re: [HCP-Users] Processing Diffusion MRI Data > > > > > > Hi Matt and Stam, > > > > I seem to still be having some sort of issue running bedpostx. I’m using: > > > > bedpostx <subject_directory> --cnonlinear --rician --model=2 --nf=3. > > > > (The –g option is not available in my version of FSL (5.0.5). I am on a > cluster environment.) > > > > So when I run this line of code, I get the following error: > > ** ERROR: nifti_convert_nhdr2nim: bad dim[0] > > ** ERROR: (nifti_image_read): cannot create nifty image from header ‘data’ > > ** ERROR: nifti_image_open (data): bad header info > > ** ERROR: failed to open file data > > > > The error seems to be with data.nii.gz, but is there error actually > occurring because I am not feeding in grad_dev.nii.gz via the –g option? Do > I need a newer version of FSL? > > > > Thanks, > > Derek Archer > > > > *From:* Stamatios Sotiropoulos [ > mailto:[email protected] > <[email protected]>] > *Sent:* Monday, February 16, 2015 12:48 PM > *To:* Archer,Derek B > *Cc:* [email protected] > *Subject:* Re: [HCP-Users] Processing Diffusion MRI Data > > > > Hi Derek > > > > Please notice that model=3 is not officially released yet in FSL. Also, > there should be a future HCP release in the near future including bedpostx > results, with the optimal parameters. > > > > Stam > > > > > > > > On 16 Feb 2015, at 16:25, Archer,Derek B <[email protected]> wrote: > > > > > > Matt, > > > Thanks for the fast reply. I will use these three inputs in bedpostX. > > > > Derek > > > > *From:* Glasser, Matthew [mailto:[email protected] > <[email protected]>] > *Sent:* Monday, February 16, 2015 11:24 AM > *To:* Archer,Derek B; [email protected] > *Subject:* Re: [HCP-Users] Processing Diffusion MRI Data > > > > Why do you need to use FLIRT/FNIRT with already registered data? As for > the others and aside from the flags you already mention, I think the > recommendation for bedpostX will include these flags: > > > > -n 3 (for 3 fibers) > > --cnonlinear > > --rician > > --model=3 > > > > Not all of this may yet be available in the public release (you might need > to use --model=2 for now). > > > > Peace, > > > > Matt. > > > > *From: *<Archer>, Derek B <[email protected]> > *Date: *Monday, February 16, 2015 at 9:16 AM > *To: *"[email protected]" <[email protected]> > *Subject: *[HCP-Users] Processing Diffusion MRI Data > > > > Hello, > > > > I am trying to analyze the preprocessed diffusion data, however, I am > having some difficulties. Is there some documentation that I could look at > that outlines how to use bedpostx, FLIRT/FNIRT, DTIFIT and probtrackx with > this data? > > > > From what I’ve found in the archives, I need to do the following: > > > > Bedpostx: include the –g option > > FLIRT/FNIRT: do I include any extra options here? > > DTIFIT: include the –gradnonlin option > > > > Is this all that needs to be done to have this data ready for > tractography? Or are there steps that I have missed? > > > > Please let me know if there is any documentation I have failed to locate. > > > > Thanks, > > Derek Archer > > > > > > > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > > ------------------------------ > > The materials in this message are private and may contain Protected > Healthcare Information or other information of a sensitive nature. If you > are not the intended recipient, be advised that any unauthorized use, > disclosure, copying or the taking of any action in reliance on the contents > of this information is strictly prohibited. If you have received this email > in error, please immediately notify the sender via telephone or return mail. > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
